Implicit Taylor methods for stiff stochastic differential equations (Q5939898)

Three implicit numerical methods are presented for approximating the solution of a stiff Itô stochastic differential equation of the form \[ dy(t)= f(y(t)) dt+ \sum^d_{j=1} g_i(y(t)) dW_j(t),\quad y(t_0)= y_0, \] where \(W_j(t)\), \(j= 1,\dots, d\), are independent Wiener processes. The first, called the implicit Euler-Taylor method (IET), is shown to have strong convergence of order .5 and a superior region of mean square stability when compared to its corresponding explicit and semi-explicit methods. The second called the implicit Milstein-Taylor method (IMT), is shown to have strong convergence of order 1.0 and a superior region of mean square stability when compared to its corresponding explicit and semi-explicit method, but a slightly inferior region as a compared to the IET method. The third, called the implicit 1.5 Taylor, is shown to have strong convergence of order 1.5, but an inferior region of mean square stability when compared to its corresponding semi-explicit method and a very inferior region when compared to the IET and IMT methods. The paper concludes by giving numerical results for two test equations.